Java 调用 shell 脚本详解

这一年的项目中，有大量的场景需要Java 进程调用 Linux的bash shell 脚本实现相关功能。

从之前的项目中拷贝的相关模块和网上的例子来看，有个别的“陷阱”造成调用shell 脚本在某些特殊的场景下，有一些奇奇怪怪的bug。

大家且听我一一道来。

先看看网上搜索到的例子：

package someTest;

import java.io.BufferedReader;
import java.io.IOException;
import java.io.InputStreamReader;

public class ShellTest {

    public static void main(String[] args) {
        InputStreamReader stdISR = null;
        InputStreamReader errISR = null;
        Process process = null;
        String command = "/home/Lance/workspace/someTest/testbash.sh";
        try {
            process = Runtime.getRuntime().exec(command);
            int exitValue = process.waitFor();

            String line = null;

            stdISR = new InputStreamReader(process.getInputStream());
            BufferedReader stdBR = new BufferedReader(stdISR);
            while ((line = stdBR.readLine()) != null) {
                System.out.println("STD line:" + line);
            }

            errISR = new InputStreamReader(process.getErrorStream());
            BufferedReader errBR = new BufferedReader(errISR);
            while ((line = errBR.readLine()) != null) {
                System.out.println("ERR line:" + line);
            }
        } catch (IOException | InterruptedException e) {
            e.printStackTrace();
        } finally {
            try {
                if (stdISR != null) {
                    stdISR.close();
                }
                if (errISR != null) {
                    errISR.close();
                }
                if (process != null) {
                    process.destroy();
                }
            } catch (IOException e) {
                System.out.println("正式执行命令：" + command + "有IO异常");
            }
        }
    }
}

testbash.sh

#!/bin/bash

echo `pwd`

输出结果为：

STD line:/home/Lance/workspace/someTest

Java在执行Runtime.getRuntime().exec(command)之后，Linux会创建一个进程，该进程与JVM进程建立三个管道连接，标准输入流、标准输出流、标准错误流。

上述代码，依次读取标准输出流和标准错误流，在shell给出“退出信号”后，做了相应的清理工作。

对于一般场景来说，这段代码可以凑合用了。但是，在实际场景中，会有以下几个“陷阱”。

一. 当标准输出流或标准错误流非常庞大的时候，会出现调用waitFor方法卡死的bug。

真实的环境中，当标准输出在10000行左右的时候，就会出现卡死的情况。

原因分析：假设linux进程不断向标准输出流和标准错误流写数据，而JVM却不读取，数据会暂存在linux缓存区，当缓存区存满之后导致该进程无法继续写数据，会僵死，导致java进程会卡死在waitFor()处，永远无法结束。

解决方式：由于标准输出和错误输出都会向Linux缓存区写数据，而脚本如何输出这两种流是Java端不能确定的。为了不让shell脚本的子进程卡死，这两种输出需要分别读取，而且不能互相影响。所以必须新开两个线程来进行读取。

我开始的实现如下：

package someTest;

import java.io.BufferedReader;
import java.io.IOException;
import java.io.InputStream;
import java.io.InputStreamReader;
import java.util.LinkedList;
import java.util.List;

public class CommandStreamGobbler extends Thread {

    private InputStream is;

    private String command;

    private String prefix = "";

    private boolean readFinish = false;

    private boolean ready = false;

    private List<String> infoList = new LinkedList<String>();

    CommandStreamGobbler(InputStream is, String command, String prefix) {
        this.is = is;
        this.command = command;
        this.prefix = prefix;
    }

    public void run() {
        InputStreamReader isr = null;
        try {
            isr = new InputStreamReader(is);
            BufferedReader br = new BufferedReader(isr);
            String line = null;
            ready = true;
            while ((line = br.readLine()) != null) {
                infoList.add(line);
                System.out.println(prefix + " line: " + line);
            }
        } catch (IOException ioe) {
            System.out.println("正式执行命令：" + command + "有IO异常");
        } finally {
            try {
                if (isr != null) {
                    isr.close();
                }
            } catch (IOException ioe) {
                System.out.println("正式执行命令：" + command + "有IO异常");
            }
            readFinish = true;
        }
    }

    public InputStream getIs() {
        return is;
    }

    public String getCommand() {
        return command;
    }

    public boolean isReadFinish() {
        return readFinish;
    }

    public boolean isReady() {
        return ready;
    }

    public List<String> getInfoList() {
        return infoList;
    }
}

package someTest;

import java.io.IOException;
import java.io.InputStreamReader;

public class ShellTest {

    public static void main(String[] args) {
        InputStreamReader stdISR = null;
        InputStreamReader errISR = null;
        Process process = null;
        String command = "/home/Lance/workspace/someTest/testbash.sh";
        try {
            process = Runtime.getRuntime().exec(command);

            CommandStreamGobbler errorGobbler = new CommandStreamGobbler(process.getErrorStream(), command, "ERR");
            CommandStreamGobbler outputGobbler = new CommandStreamGobbler(process.getInputStream(), command, "STD");

            errorGobbler.start();
            // 必须先等待错误输出ready再建立标准输出
            while (!errorGobbler.isReady()) {
                Thread.sleep(10);
            }
            outputGobbler.start();
            while (!outputGobbler.isReady()) {
                Thread.sleep(10);
            }

            int exitValue = process.waitFor();
        } catch (IOException | InterruptedException e) {
            e.printStackTrace();
        } finally {
            try {
                if (stdISR != null) {
                    stdISR.close();
                }
                if (errISR != null) {
                    errISR.close();
                }
                if (process != null) {
                    process.destroy();
                }
            } catch (IOException e) {
                System.out.println("正式执行命令：" + command + "有IO异常");
            }
        }
    }
}

到此为止，解决了Java卡死shell脚本的情况。再说说，第二种可能。

二. 由于shell脚本的编写问题，当其自身出现僵死的情况，上述代码出现Java代码被僵死的Shell脚本阻塞住的情况。

原因分析：由于shell脚本也是人写的，难免会出现失误。在Java调用shell脚本时，无论是Debug场景还是生产环境，都发生过shell脚本意外僵死反过来卡死Java相关线程的情况。典型的表现为：shell脚本长时间运行，标准输出和错误输出没有任何输出（包括结束符），操作系统显示shell脚本在正常运行或僵死，没有退出信号。

解决方式：上述代码中，至少有三处会导致线程阻塞，包括标准输出和错误输出这线程的BufferedReader的readline方法，以及Process的waitFor方法。解决这个问题的核心有两个，1.避免任何Java线程被阻塞住，因为一旦被IO阻塞住，线程将处于内核态，主线程没有任何办法强制结束相关子线程。2.添加一个简单的超时机制，超时后回收相应的线程资源，并结束调用过程。

演示代码中，我改写了testshell.sh，写一个没有任何输出的死循环模拟shell卡死的情况。

#!/bin/bash

while true;do
    a=1
    sleep 0.1
done

package someTest;

import java.io.BufferedReader;
import java.io.IOException;
import java.io.InputStream;
import java.io.InputStreamReader;
import java.util.LinkedList;
import java.util.List;

public class CommandStreamGobbler extends Thread {

    private InputStream is;

    private String command;

    private String prefix = "";

    private boolean readFinish = false;

    private boolean ready = false;

        // 命令执行结果,0:执行中 1:超时 2:执行完成
        private int commandResult = 0;

    private List<String> infoList = new LinkedList<String>();

    CommandStreamGobbler(InputStream is, String command, String prefix) {
        this.is = is;
        this.command = command;
        this.prefix = prefix;
    }

    public void run() {
        InputStreamReader isr = null;
        BufferedReader br = null;
        try {
            isr = new InputStreamReader(is);
            br = new BufferedReader(isr);
            String line = null;
            ready = true;
            while (commandResult != 1) {
                if (br.ready() || commandResult == 2) {
                                    if ((line = br.readLine()) != null) {
                                        infoList.add(line);
                                    } else {
                                        break;
                                    }
                                } else {
                                    Thread.sleep(100);
                                }
            }
        } catch (IOException | InterruptedException ioe) {
            System.out.println("正式执行命令：" + command + "有IO异常");
        } finally {
            try {
                if (br != null) {
                    br.close();
                }
                if (isr != null) {
                    isr.close();
                }
            } catch (IOException ioe) {
                System.out.println("正式执行命令：" + command + "有IO异常");
            }
            readFinish = true;
        }
    }

    public InputStream getIs() {
        return is;
    }

    public String getCommand() {
        return command;
    }

    public boolean isReadFinish() {
        return readFinish;
    }

    public boolean isReady() {
        return ready;
    }

    public List<String> getInfoList() {
        return infoList;
    }

    public void setTimeout(int timeout) {
        this.commandResult = timeout;
    }
}

package someTest;

public class CommandWaitForThread extends Thread {

    private Process process;
    private boolean finish = false;
    private int exitValue = -1;

    public CommandWaitForThread(Process process) {
        this.process = process;
    }

    public void run() {
        try {
            this.exitValue = process.waitFor();
        } catch (InterruptedException e) {
            e.printStackTrace();
        } finally {
            finish = true;
        }
    }

    public boolean isFinish() {
        return finish;
    }

    public void setFinish(boolean finish) {
        this.finish = finish;
    }

    public int getExitValue() {
        return exitValue;
    }

}

package someTest;

import java.io.IOException;
import java.io.InputStreamReader;
import java.util.Date;

public class ShellTest {

    public static void main(String[] args) {
        InputStreamReader stdISR = null;
        InputStreamReader errISR = null;
        Process process = null;
        String command = "/home/Lance/workspace/someTest/testbash.sh";
        long timeout = 10 * 1000;
        try {
            process = Runtime.getRuntime().exec(command);

            CommandStreamGobbler errorGobbler = new CommandStreamGobbler(process.getErrorStream(), command, "ERR");
            CommandStreamGobbler outputGobbler = new CommandStreamGobbler(process.getInputStream(), command, "STD");

            errorGobbler.start();
            // 必须先等待错误输出ready再建立标准输出
            while (!errorGobbler.isReady()) {
                Thread.sleep(10);
            }
            outputGobbler.start();
            while (!outputGobbler.isReady()) {
                Thread.sleep(10);
            }

            CommandWaitForThread commandThread = new CommandWaitForThread(process);
            commandThread.start();

            long commandTime = new Date().getTime();
            long nowTime = new Date().getTime();
            boolean timeoutFlag = false;
            while (!commandIsFinish(commandThread, errorGobbler, outputGobbler)) {
                if (nowTime - commandTime > timeout) {
                    timeoutFlag = true;
                    break;
                } else {
                    Thread.sleep(100);
                    nowTime = new Date().getTime();
                }
            }
            if (timeoutFlag) {
                // 命令超时
                errorGobbler.setTimeout(1);
                outputGobbler.setTimeout(1);
                System.out.println("正式执行命令：" + command + "超时");
            }else {
                // 命令执行完成
                errorGobbler.setTimeout(2);
                outputGobbler.setTimeout(2);
                        }

            while (true) {
                if (errorGobbler.isReadFinish() && outputGobbler.isReadFinish()) {
                    break;
                }
                Thread.sleep(10);
            }
        } catch (IOException | InterruptedException e) {
            e.printStackTrace();
        } finally {
            if (process != null) {
                process.destroy();
            }
        }
       }

    private boolean commandIsFinish(CommandWaitForThread commandThread, CommandStreamGobbler errorGobbler, CommandStreamGobbler outputGobbler) {
        if (commandThread != null) {
            return commandThread.isFinish();
        } else {
            return (errorGobbler.isReadFinish() && outputGobbler.isReadFinish());
        }
    }
}

在以上的代码中，为了防止线程被阻塞，要点如下：

1. 在CommandStreamGobbler里，bufferedReader在readLine()之前，先用ready()看一下当前缓冲区的情况，请特别注意ready()描述，这个方法是非阻塞的。

boolean java.io.BufferedReader.ready() throws IOException

Tells whether this stream is ready to be read. A buffered character stream is ready if the buffer is not empty, or if the underlying character stream is ready.

Returns:
True if the next read() is guaranteed not to block for input, false otherwise. Note that returning false does not guarantee that the next read will block.

2.在一个新线程commandThread中，调用process对象的waitFor()从而避免主线程卡死，主线程的最后会执行finally块中的process.destory()保证commandThread正常退出。

以上的两点改进，保证了Java在调用shell脚本过程互不被对方卡死的机制。

三.在执行shell脚本过程中，可能会添加参数，通常在终端中，我们使用“ ”（空格）把参数隔开。

为了区分空格是作为参数分隔符，还是参数的一部分。调用exec方法有特别的注意事项。

String command = "/home/Lance/workspace/someTest/testbash.sh 'hello world'";

process = Runtime.getRuntime().exec(command);

等价于

List<String> commandList = new LinkedList<String>();
commandList.add("/home/Lance/workspace/someTest/testbash.sh");
commandList.add("hello world");
String[] commands = new String[commandList.size()];
for (int i = 0; i < commandList.size(); i++) {
    commands[i] = commandList.get(i);
}

process = Runtime.getRuntime().exec(commands);

好了，今天介绍到这里。